The Intermittent That Only Appeared on Laundry Day

Handle the hazards before the investigation

Stored water at a sanitizing temperature will scald. Do not open, break or probe a hot line while it is under pressure and hot: close the isolation valve, relieve the pressure at a drain point, and let the line cool before any joint comes apart. If the equipment at the end of the run is electrically powered and standing on a wet floor, de-energize at the disconnect, lock and tag it, and verify dead before a probe touches a terminal, which is the duty at 29 CFR 1910.333(b)(2) for work on electric circuit parts; prove dead with the live-dead-live sequence in NFPA 70E-2021, 120.5.

One more, because it is the tempting shortcut on this exact fault: raising a water heater's stored temperature to force more heat down a long run raises the scald risk at every fixture served by that heater, so it is not an option unless thermostatic mixing is provided downstream at the fixtures that people touch. Read that as a gate, not a warning. Without the mixing, the setpoint does not move.

The call: a fault that has already stopped

A small commercial tenant has a machine that runs a heated sanitizing cycle at the far end of the suite. For about a year it failed its temperature check, sometimes. Three visits across that year, three write-ups, one part replaced, no resolution. Six weeks before this call it simply stopped happening, and the tenant now wants to know whether it is fixed or whether it is waiting.

There is nothing to reproduce on arrival, which rules out the normal approach entirely. The only evidence still in existence is written down somewhere, so the investigation runs backwards through the paperwork before it runs forwards through the plumbing.

What the three tickets actually said

Pulled from the shop's own history, in order:

• Visit one, about 14 months ago. "Customer reports machine failing temp check intermittently. Checked hot water at sink, hot. Flushed water heater. Unable to reproduce." On site 1.5 hours.
• Visit two, about 9 months ago. "Same complaint. Replaced heating element in machine. Ran cycle, passed." On site 2.0 hours.
• Visit three, about 4 months ago. "Recurring. All readings normal. Advised monitor and call when it happens." On site 1.0 hours.

Four and a half hours of labor across three visits, one part, and a machine that kept failing. Read those three entries as data about the shop rather than about the plumbing and a pattern comes out immediately: every visit tested a system that was behaving.

The three gaps, and what each one cost

The gaps in a record are findings. These three are the whole case.

No ticket records a time of day. All three carry a date and none carries a clock time. Dispatch records filled that in: the three visits started at 9:10 am, 8:45 am and 10:05 am. Three morning visits, on a complaint that the tenant would later place squarely in the afternoon. Nobody chose morning for a reason. Morning is simply where a dispatcher puts a "could not reproduce" recall, because it is the slot nobody else wants.

No ticket records what else was running. Not one of the three notes any other water draw, appliance or piece of equipment operating during the visit or during the reported failures. That single missing field is what kept this fault alive for a year, and it costs one question to fill.

No ticket records a temperature at the machine's inlet. Visit one measured hot water at a sink, which is a different branch, close to the heater, and proves nothing about a long run. Visit three wrote "all readings normal" without naming a single reading, which is the most expensive sentence in the file: the next tech cannot tell whether the right measurement was taken and passed, or never taken at all.

Visit two is worth its own note. A part went in, the cycle passed, and everyone moved on. That cycle passed at mid-morning, when the fault does not occur, so the part had no chance to be tested against the failure condition. The element is now a variable in the system rather than a constant, and nobody wrote down what the old one measured before it came out.

Rebuilding the timeline from the tenant's own records

The tenant keeps a laundry log for their own compliance reasons, listing wash loads by date and time. They also remember which days the machine failed, roughly, and their staff scheduling is on a calendar.

Cross-referencing the three sources gives a timeline the shop never had:

• Reported failures cluster on two weekdays. The laundry log shows heavy loads on those same two weekdays.
• Laundry runs in a block from roughly 1 pm to 4 pm, because that is when the staff who does it is on shift.
• Six weeks ago the tenant changed the staff schedule, and laundry moved to a morning block. The failures stopped that same week.

That last line is the one that matters. The fault did not get fixed, it got avoided, and it is one schedule change away from returning. Note also the grim symmetry: laundry now runs in the mornings, which is exactly when the shop visits, so if the tenant had called during those six weeks the fault would finally have been reproducible on a morning visit. Timing had been against everyone the entire year.

The measurement nobody had taken

With the tenant's permission, the pair is run deliberately: the laundry load and the machine's heated cycle together, in the afternoon, with a thermometer at the machine's inlet. All temperatures in degrees Fahrenheit, all readings illustrative of the method rather than universal values.

• Heater outlet, measured: 140.
• Machine inlet with no other draw: 128.
• Machine inlet during the laundry draw: 111.
• Machine data plate minimum inlet temperature: 120.

The arithmetic does the work. The total headroom between the heater outlet and the machine's requirement is 20 degrees. The run itself consumes 12 of those before any other demand exists, leaving a standing margin of 8 degrees. The concurrent laundry draw then takes 17 more, landing at 111, which is 9 degrees below the minimum the machine needs.

A single-actor test at that same inlet returns 128, which clears the 120 minimum by 8 degrees. That reading is correct, it is documented, and it exonerates a system that fails every laundry afternoon. This is what "all readings normal" looked like from the inside.

Reading the two contributors apart

There are two independent contributors and the arithmetic tells you which fix actually works, which is the part most quotes get wrong.

The distance contributor costs 12 degrees, standing, with nothing else running. That is a long uninsulated run with a lot of surface area and a lot of dwell time between draws. It is a real defect and it is the reason the margin was only 8 degrees to begin with.

The concurrency contributor costs 17 degrees on top, because the laundry draw diverts supply and pulls the far branch's delivered temperature down while the heater is recovering.

Now test the tempting single fix. Suppose insulating and rerouting recovers 7 of the 12 degrees lost to the run: the standing inlet reads 135, and the same 17-degree concurrent loss lands it at 118. Still 2 degrees under the minimum. The run improvement alone does not fix this, and a shop that sells only that is selling a callback. In fairness the 17-degree figure would shrink somewhat on a better-insulated line, since a warmer standing line loses less during a draw, so treat 118 as a boundary rather than a prediction. It is close enough to the limit that no honest quote would rest on it.

Eliminating the concurrency, by contrast, returns the inlet to the 128 standing reading, which clears the minimum by 8 degrees with the run exactly as it is today. That is the fix that works on arithmetic available right now: separate the demands, either by dedicating supply to the machine or by scheduling the two so they do not overlap.

The honest recommendation names both, in order. Separate the demands to make the machine reliable. Improve the run to rebuild the 12 degrees of margin that make the whole system tolerant of the next new load somebody adds. Sold as one item, the second one looks like padding. Sold with the numbers above, it is obviously the difference between a system that passes and a system that passes with room.

Proving it before quoting

Reproduction is not optional on a fault this old, because the tenant has been told three times that nothing is wrong and will not accept a fourth theory.

The demonstration is three readings taken in front of them, in one visit: inlet at 128 with nothing else drawing, inlet at 111 during a deliberate laundry draw, inlet back at 128 within a few minutes of the draw ending. Bad with both, fine with one, and it recovers when the second demand stops. That takes under half an hour and it converts a year of disputed history into an observed fact.

Repeat the pair once. A single occurrence is an event, and this customer has earned a demonstration rather than a claim.

What went on the record this time

The write-up carries what the previous three did not, because the next person to touch this suite should not have to redo any of it:

• The three measured temperatures, each with its measurement point named, plus the data-plate minimum they are being judged against.
• The concurrent condition, in the customer's own operating terms: heavy laundry block, afternoon, two weekdays.
• The fact that the failure is currently masked by a schedule change made six weeks ago, and that it returns if laundry moves back to afternoons.
• The status of the element replaced at visit two: installed nine months ago, never tested against the failure condition, verified functional now, no record of the original part's condition.

That last bullet is the kind of entry that feels like admitting a mess. It is worth more than the other three combined, because it is what stops the next tech from treating a nine-month-old part as a known-good and spending an hour proving something that is already written down.

References

• 29 CFR 1910.333(b)(2), OSHA electrical safe work practices, for de-energizing and locking or tagging before work on electric circuit parts near wet floors
• NFPA 70E-2021, 120.5, the live-dead-live process for establishing an electrically safe work condition
• Manufacturer data plate and service literature for minimum inlet temperature and cycle requirements on the specific machine
• Trade-standard practice for scald protection and thermostatic mixing where stored water temperature is raised
• See related: The Fault That Only Exists Under Concurrent Demand; How to Test Two Systems Running at Once; How to Identify a Supply-Caused Fault Before Replacing Parts